Does anybody know of a word2vec model that is trained on object definitions? Perhaps something trained on an encyclopedia? I can’t seem to find anything online.<\/p>\n
My ideal scenario would be that it finds similarities between, say, “rollercoaster”, and its constituent parts (metal, tracks, moving fast, speed), etc.<\/p>\n
Or between “saturn” and (rings, space, stars, gas, yellow, huge)<\/p>\n
It’s a little more complex than the above examples, but I’m pretty solid on the approach, so I’ve simplified it for ease.<\/p>\n
If there are none trained on encylopdia, would Wikipedia be a suitable dataset for this kind of use case? <\/p>\n
(Before anyone says the obvious; I know that Wikipedia is an “online encyclopedia,” but as you all know, it goes way further than that. There are wiki pages for all sorts of games, events like natural disasters, etc, and I’m worried that those might taint the data pool.)<\/p>\n<\/div>\n
submitted by \/u\/notquitehuman_ <\/a>
[link]<\/a><\/span> [comments]<\/a><\/span><\/p>
